Reading contributions to a certain audio forum recently, it became apparent to me that some audiophiles have rather naive expectations when it comes to the ability of science to drive the development of audio technology.
Common misapprehensions are that science occurs automatically when the scientific method is followed, and that science proves things with certainty.
People rarely understand that science has limits. There are are some fields of enquiry where science is just not applicable. It is possible to follow the scientific methodology to the letter, and yet for the whole thing to simply be non-science. The applicability, or otherwise, of science is not always obvious.
Even if the science is more-or-less valid, there is often a blatant disregard for the ‘terms and conditions’ that accompany the findings. So, if science purportedly shows that “phase is inaudible”, we should examine exactly what the experiments covered before accepting it at face value.
Although a scientific method may have been followed, with full double blind protocol, the would-be scientist’s prejudices can be woven into the results merely through the existence of the experiment itself. If we are testing for the audibility of phase distortion, but choose to do so only for mono signals, we may end up confirming what we wanted the result to be in the first place. The justification for using mono is made very un-scientifically by spinning a few words out of thin air, but the arbitrariness of this is lost when quoting the final conclusion: “Phase is inaudible”. On the back of these ‘findings’ an industry can carry on its merry way ignoring a major distortion of the signal.
Listening tests are a red herring. In audio, they are seen as scientific polyfilla capable of filling the gap between objective measurements and the touchy-feely world of human perception. Because listening tests have the patina of science, they are believed to be capable of resolving questions of superiority between differing technologies that ordinary science wouldn’t touch. The idea is that although listening tests may not tell us why things work, they can give us scientific validation that they do, anyway. I think this is just pseudo-science. If a claim is based on listening tests alone then all bets are off, as listening tests can be ‘gamed’ in so many ways. As alluded to above, even if the actual test is carried out with rigorous adherence to scientific method, there are so many aspects that can only be justified with hand-waving i.e. the very premise of the test, the selection of listeners, the choice of listening material, what the listeners are listening for and how the question is worded (is it a “difference” or a “preference”?), and so on. The ‘device under test’ cannot be isolated from the characteristics of the remainder of the listening hardware. Can we really tell whether feedback in amplifiers is a Bad Thing (a typical audiophile preoccupation) if the transducers are decidedly nonlinear? Sure, we can preface the ‘findings’ with details of which speakers were used and assert that these represent “typical high quality domestic loudspeakers”, but that’s just words.
I think that hi-fi is a lot simpler than the standard orthodoxy maintains. Even people who like to think of themselves as objectivists have been seduced into believing that there are mysterious factors at work, beyond simple fidelity to the recorded signal. Either they don’t believe that it is possible to recreate the signal faithfully and therefore we must introduce compensating mechanisms to mask the deficiencies, or they believe that faithful reproduction is not enough and that some extra spice must be added to make the result believeable. This may take the form of adding “euphonic distortion” or other deliberate colourations, and can justify the use of valves, vinyl and passive crossovers despite their apparent inferiority in measurements. They believe that listening test-based experiments can harden these vague notions into scientific certainty.
This does not imply that I believe that objective measurements tell all. As with listening tests, relying on measurements to steer the design process is a road to nowhere. Unlike some measurements-oriented people, I am happy to accept that audio equipment can contribute distortions that evade the usual tests – I am sure I could deliberately design some hardware to do that – so without looking inside the black box we can never be sure that our tests have caught all peculiarities. My view is that measurements can only be used to confirm that a design is working, rather than proving that it is a good design in the first place.
In reality, hi-fi can never stray far from the central necessity of reproducing the signal faithfully (and even if there was some magical formula that could make it sound more real than it does, it could be applied with an effects box rather than hoping to achieve it indirectly within the amplifier or through ‘vinyl-isation’). The real magic occurs when everything is done to make the whole system linear – not simply substituting a super-accurate DAC for vinyl, and carrying on with ye olde worlde valve amplifier and loudspeakers. It doesn’t need listening tests, and it doesn’t need science; only engineering.